Fabric Is Just Plain Unreliable, and Microsoft's Hiding It | Hacker News Reader

Fabric Is Just Plain Unreliable, and Microsoft's Hiding It

brentozar.com

117 pointsko_pivot12 comments

Yep, this is common - especially for monday mornings where jobs failed on Thursday/Friday. Said department would run to IT/Cloud computing team to see what was going on - and it always led to an instant ticket because nothing was wrong and the status page did not reflect accurately the service outage.

Reddit usually was the only place that reported an outage, the o365 support twitter is dead now, and the staus page per tenant is reporting is unreliable because they are manually updated.

AFAIK the SLA for fabric/PowerBI does fall under M365 if you have it.

Just in case someone else also had no idea what Microsoft Fabric is: it seems to be some sort of integrated platform for data processing and analysis with AI features (because of course there's AI)

It’s way more than that. It’s an end to end data integration, data lake and pipeline management platform.

Sure. I have no idea what any of those words mean though because I've never worked with a platform like that. The way I described it is how I understood it as a person who has no idea what it is.

Microsoft seems to just be expanding rapidly, and not worrying too much about feature parity, compatibility, or reliability. Why did our logic app fail last night? No reason, just Azure hiccups. Why doesn't the Sentinel data connector work? Whoever maintains it doesn't care. etc.

Fabric will be great in 5 years, but right now it tends to be unreliable, unergonomic, and surprisingly expensive.

As a data consultant, I will help you make use if it, but I will not advise you to adopt it if you can avoid it.

I looked at Fabric a year ago and you're right about it being expensive. That's part of the reason (another being that most of it was just in experimental state at that time) I decided to not use it for production. I'll review it again when MS finalized most of the features, and when (hopefully) they reduce the pricing.

Author here - whoa, didn’t expect this to hit HN. Here for any questions, but I think the post speaks for itself.

I wonder if this is a difference in mindset of OLAP vs OLTP. I have worked with OLAP teams where they were dropping millions of records, and the solution was usually just restream the data and shrug. But if any service backed by OLTP was dropping writes, its all-hands-on-deck incident with detailed postmortems and exec meetings.

If you don't own the source system, then it's much harder to be incentivized to care.

Also the better data teams do postmortems and exec meetings.

(I haven't been hands-on with Fabric (I've used ADF and PowerBI pre-Fabric), and I work for a company that sometimes works with Fabric competitors, so I am biased.)

From our experience with customers, Fabric feels like a rebrand of existing Azure services, plus a few new beta services that Microsoft is aggressively selling to their customers as fully baked. I've heard two customer stories this year where they migrated ETL to Fabric and then had to scramble to migrate again because of latency or reliability issues.

I suspect Fabric will be pretty dang good a few years from now, and that it's probably a great option today for users who accurately understand its strengths and weaknesses, but I think the way some MS sales teams are selling the platform to customers right now is dishonest and harmful.

> Fabric feels like a rebrand of existing Azure services, plus a few new beta services that Microsoft is aggressively selling to their customers as fully baked

Couldn't agree more. To me, Synapse and Data Factory (ADF) are somewhat overlapping services too (sure Synapse can do some stuff like running SQL against blob storage files). I wonder what would happen to Synapse and ADF when Fabric becomes fully functional. I hope they don't kill ADF though.

Reply on news.ycombinator.com